This improves the AI behavior by choosing the stances non-randomly:
* Breakthrough will be used if the base is expected to be capturable and
the coalition outnumbers the enemy by 20%.
* Elimination will be used if the coalition has at least as many units
as the enemy.
* Defensive will be used if the coalition has at least half as many
units as the enemy.
* Retreat will be used if the coalition is significantly outnumbers.
This also exposes the option to the player.
An HTN (https://en.wikipedia.org/wiki/Hierarchical_task_network) is
similar to a decision tree, but it is able to reset to an earlier stage
if a subtask fails and tasks are able to account for the changes in
world state caused by earlier tasks.
Currently this just uses exactly the same strategy as before so we can
prove the system, but it should make it simpler to improve on task
planning.
This doesn't currently interact very well with the auto purchase since
the procurer might by aircraft that don't have pilots available. That
should be fixed, but for the short term we should just default to not
enabling this new feature.
If these orders can't be fulfilled for the faction it will prevent the
faction from ordering any non-reserve aircraft since transports are
given priority after reserve missions, and they'll never be fulfillable.
As such, no non-reserve aircraft will ever be purchased for factions
without transport aircraft.
Factions without transport aircraft are screwed in other ways, but this
will fix their air planning for campaigns that aren't dependent on
airlift.
We were doignt his for escorts, but now that we quit planning as soon as
we find an unplannable mission (to save money for higher priority
missions), if we hit an early unplannable mission like BARCAP no other
missions wil be planned.
Maybe fixes https://github.com/dcs-liberation/dcs_liberation/issues/1228
This is an attempt to remove a lot of our supposedly unnecessary error
handling. Every aircraft should have a price, a description, a name,
etc; and none of those should require carrying around the faction's
country as context.
This moves all the data for aircraft into yaml files (only one converted
here as an example). Most of the "extended unit info" isn't actually
being read yet.
To replace the renaming of units based on the county, we instead
generate multiple types of each unit when necessary. The CF-18 is just
as much a first-class type as the F/A-18 is.
This doesn't work in its current state because it does break all the
existing names for aircraft that are used in the faction and squadron
files, and we no longer let those errors go as a warning. It will be an
annoying one time switch, but it allows us to define the names that get
used in these files instead of being sensitive to changes as they happen
in pydcs, and allows faction designers to specifically choose, for
example, the Su-22 instead of the Su-17.
One thing not handled by this is aircraft task capability. This is
because the lists in ai_flight_planner_db.py are a priority list, and to
move it out to a yaml file we'd need to assign a weight to it that would
be used to stack rank each aircraft. That's doable, but it makes it much
more difficult to see the ordering of aircraft at a glance, and much
more annoying to move aircraft around in the priority list. I don't
think this is worth doing, and the priority lists will remain in their
own separate lists.
This includes the converted I used to convert all the old unit info and
factions to the new format. This doesn't need to live long, but we may
want to reuse it in the future so we want it in the version history.
This Pull Request lets users plan Tanker flights.
Features:
- Introduction of `Refueling` flight type.
- Tankers can be purchased at airbases and carriers.
- Tankers get planned by AI.
- Tankers are planned from airbases and at aircraft carriers.
- Tankers aim to be at high, fast, and 70 miles from the nearest threat.
(A10s won't be able to tank)
- Tankers racetrack orbit for one hour.
- Optional Tickbox to enable legacy tankers.
- S-3B Tanker added to factions.
- KC-130 MPRS added to factions.
- Kneeboard shows planned tankers, their tacans, and radios.
Limitations:
- AI doesn't know whether to plan probe and drogue or boom refueling
tankers.
- User can't choose tanker speed. Heavily loaded aircraft may have
trouble.
- User can't choose tanker altitude. A-10s will not make it to high
altitude.
Problems:
- Tanker callsigns do not increment, see attached image. (Investigated:
Need to use `FlyingType.callsign_dict`, instead of just
`FlyingType.callsign`. This seems like it might be significant work
to do.).
- Having a flight of two or more tankers only spawns one tanker.
- Let me know if you have a solution, or feel free to commit one.
https://user-images.githubusercontent.com/74509817/120909602-d7bc3680-c633-11eb-80d7-eccd4e095770.png
Missions with very large numbers of packages and short mission windows
would raise an exception here because we couldn't schedule more
frequently than once a minute. Switch to using seconds instead of
minutes to avoid that problem. If there are more packages than there are
seconds in the mission the game is broken for other reasons.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/1154
The usual symptom here was the game breaking when a carrier is
destroyed. The carrier would no longer be operational but missions would
be assigned there that could not generate flight plans.
Target the air defenses whose *threat ranges* come closest to friendly
bases rather than the closest sites themselves. In other words, the
SA-10 that is 5 miles behind the SA-6 will now be the priority.
This also treats EWRs a bit differently. If they are not protected by a
SAM their detection range will be used for determining their "threat"
range. Otherwise a heuristic is used to determine whether or not they
can be safely attacked without encroaching on the covering SAM.
We shouldn't consider the non-escorted parts of the flight path when
checking for threats to determine if escorts should be used or not,
since escorts can't help in those areas anyway. This was causing escorts
to be overly requested since the bullseye is now a part of the
"flight plan", but could have also triggered for divert waypoints, or
for aircraft taking off in a retreat from a threatened location.
We don't need to include a SEAD flight in missions against EWRs or SAMs
that no longer have a radar.
Also plan DEAD missions against air defenses that have no radars.
Previously we would never finish killing launcher only sites (which
cannot defend any more, but are cheaper to return to working order than
a fully destroyed site) nor would we plan DEAD against IR SAMs or AAA.
The priority list was guiding the purchase decision which largely meant
that this was working correctly, but there were suboptimal cases where
the list was being taken in FIFO order by purchased type. This fixes the
search to be locally optimal, although this does still mean that a worse
but closer aircraft will be chosen over a better but slightly farther
away aircraft. We'd need to have a quality vs distance rating to do
better.
Fixes https://github.com/dcs-liberation/dcs_liberation/issues/755
A was intended to be the blue point and B was intended to be the red
point. Make this a part of the name so that's clear, and clean up
related code to keep that reliable.
The simple form of this works, but without the multi-mode routing it'll
only get used when the final destination is a port with a link to a port
with a factory.
These also aren't targetable or simulated yet.
https://github.com/Khopa/dcs_liberation/issues/826
This is an interim improvement since we should probably be pushing the
BARCAPs into TARCAP roles when the front line is so close. This does
regress flight pathing for anything that should route around the front
(to avoid getting shot at by SHORADS and TARCAPs), but for now it's one
or the other and this is the one everyone's complaining about.
Like the comment says this rarely has any effect due to the ordering of
flight planning and convoy creation. Could be separated, but opfor will
still not be able to target any convoys that the player creates in the
UI on that turn because they planning is done before the player can use
the UI.
Multi-turn transfers will be targetable, however.
